Two Ph.D. students advised by Todd Humphreys, a professor of aerospace engineering and engineering mechanics at The University of Texas at Austin, are one of only eight teams selected to receive a Qualcomm Innovation Fellowship for 2017. The research funded by the fellowship will focus on tracking the precise location of pedestrians and cyclists and communicating these locations to traffic management systems and automated vehicles.
Aerospace engineering student Matthew Murrian and electrical engineering student Lakshay Narula will receive $100,000 from the communications company to fund their research in UT Austin’s Radionavigation Lab. Qualcomm will provide one year of mentorship in addition to funding their research.
Nearly 50% of traffic fatalities around the world involve pedestrians or cyclists, according to the World Health Organization. There is currently a lot of work being done to develop precise localization systems for driverless vehicles, Narula says, but not much done in the way of pedestrians and cyclists. The current belief is that the quick reaction time and unwavering vigilance of self-driving cars will virtually eliminate pedestrian and cyclist fatalities.
But Narula argues that since pedestrians and cyclists move unpredictably, a self-driving car determined to avoid all accidents with these vulnerable travelers will slow to a crawl. This, Narula says, is the “crawling car conundrum.” He points out that pedestrians and cyclists who know and can communicate their precise location will reduce the risk of accidents, allowing self-driving cars to proceed at normal speeds down the roads.
Murrian and Narula are working on an innovative solution to this problem by developing precise location tracking software for pedestrians and cyclists wearing smartphones with upgraded GPS and sensor technology. Their work expands upon centimeter-accurate GPS research developed by Humphreys and his students in the Radionavigation Lab. Although precise GPS is an important factor in a device’s location accuracy, the duo says there is still room for improvement. In areas without a clear view of the sky, such as a city with tall buildings, GPS can become inaccurate.
The team believes that a “sensor fusion” approach that combines centimeter-accurate GPS with other opto-electric sensors is the solution to accurately tracking pedestrians and cyclists. The approach proposed by the team combines precise GPS with an Inertial Measurement Unit (IMU) and a visible-light camera. These sensors detect the instantaneous acceleration and rotation of the cyclist/pedestrian in a way that is complementary to GPS, which provides a global location.
“Precise GPS technology is great when it works, but when passing through a challenging environment, satellite signals may either be blocked or corrupted, and the user’s location may suddenly jump,” Murrian said. “It will jump in a very non-physical way, but we can’t know it’s non-physical unless we have the camera to tell us, ‘No, you did not just leap to the side by a meter.’ And the IMU will tell us, ‘No, you did not just rotate 30 degrees to the left.’ That’s where these sensors can come in to play for location accuracy.”
Murrian will focus on the GPS and IMU aspects of the project while Narula perfects the computer vision component. Their goal is to build a tracking device that will know a user’s location to within half a meter and the direction the user is facing to 2-3 degrees. Narula expects their research will lead to fewer traffic fatalities because it can provide autonomous cars the precise location of pedestrians and cyclists, even if the cars line-of-sight sensors cannot see them.
“Making cars autonomous and having better sensors will help reduce traffic fatalities, but if you can’t perceive the cyclist with the sensors you have, that’s where the problem is,” Narula said. “Instead of the car having to see, the pedestrian or cyclist will just provide a precise location.”
Though the team is concentrating on developing the software for the tracking device, Murrian said their prototype concept is an arm-mounted smartphone that contains an IMU and a camera facing outward. He said the most challenging part is making sure their final product will be cost effective and capable for smartphone use.
Another problem the team says they must overcome is hacking. For example, a pedestrian or cyclist could fabricate a location, forcing a car to stop when no one is actually at that location. Despite the challenges, Murrian and Narula are excited to put their problem-solving skills to the test.
“This is a very challenging problem, but that’s why we chose it,” Murrian said. “Its difficulty is what makes it exciting.”
